On Mon, 2013-08-12 at 14:30 +0100, Peter Cock wrote:
On Mon, Aug 12, 2013 at 2:18 PM, Bjoern Gruening <bjoern.gruening@gmail.com> wrote:
On Mon, 2013-08-12 at 14:13 +0100, Peter Cock wrote:
I expect that in the meantime this bug will hit other people as well (and until the Tool Shed testing is improved I'm a little worried there could be other 'broken' packages) :(
If you had a working (minimal non-Altas non-lapack) numpy tool_dependencies.xml pre-dating the template_command action it would seem sensible to upload it now to fix the numpy package on the main Tool Shed.
I would have thought one of the old revisions from the numpy package on the Test Tool Shed would work?
Yes they do. But if you could not deactivate cpu throttling it will crash and all other dependent packages will not be installed.
Ah. But at least it would work on some machines, whereas right now it will fail on all galaxy-dist installations (until the next stable release happens).
The numpy package on main under the IUC account was uploaded yesterday. I do not think any other package is using it, or?
Too late :(
These repositories on the Main Tool Shed use the numpy package directly: http://toolshed.g2.bx.psu.edu/view/biopython/package_biopython_1_61 http://toolshed.g2.bx.psu.edu/view/iuc/package_matplotlib_1_2
These all use it via the Biopython package: http://toolshed.g2.bx.psu.edu/view/bgruening/glimmer3 http://toolshed.g2.bx.psu.edu/view/peterjc/seq_select_by_id http://toolshed.g2.bx.psu.edu/view/peterjc/get_orfs_or_cdss
These will all use it via the Biopython package when next updated: http://toolshed.g2.bx.psu.edu/view/peterjc/seq_filter_by_id http://toolshed.g2.bx.psu.edu/view/peterjc/seq_select_by_id
There are probably more - Greg can perhaps do a search for declared dependencies directly even if this isn't an option via the Tool Shed website.
For these Biopython examples, they don't actually need NumPy - so we could perhaps temporarily update the Biopython package on the main Tool Shed to install without NumPy...
Maybe I should update the repository description to reflect that the newest version is needed?
Updating the description might help - but it doesn't need the newest version, it needs the not-yet-released stable version of Galaxy :(
Ahhh, right the revisions are gone. So the latest is always the one that are used. Oh oh ... I will screen through my github history and try to revert the numpy package until next stable is out. Oh that was a bad move from me. Sorry, Bjoern
Peter